TDCMR: Triplet-Based Deep Cross-Modal Retrieval for Geo-Multimedia Data

Mass multimedia data with geographical information (geo-multimedia) are collected and stored on the Internet due to wide application of location-based services (LBS). How find high-level semantic relationship between geo-multimedia construct efficient index is crucial for large-scale retrieval. To combat this challenge, paper proposes a deep cross-modal hashing framework retrieval, termed as Triplet-based Deep Cross-Modal Retrieval (TDCMR), which utilizes neural network an enhanced triplet constraint capture semantics. Besides, novel hybrid index, called TH-Quadtree, developed by combining binary hash codes quadtree support high-performance search. Extensive experiments conducted three common used benchmarks, results show superior performance proposed method.